‍Two kinds of legal work

Understanding this threat requires a useful distinction. Kevin Lee, Founding Director of the Institute for AI & Democratic Governance, argues that legal work operates on two levels, which he characterizes as syntactic and semantic.

• Syntactic work is work in which AI excels, such as document generation, pattern recognition, research compilation, and contract review. These are tasks that can be processed, matched, and produced at scale, faster and cheaper than any junior associate can do them.
• Semantic work is different because it involves exercising independent legal judgment, reflecting on consequences, understanding context, and fulfilling the fiduciary duties that define the attorney-client relationship. When a client seeks counsel, they are looking for a human being who can hold the full complexity of their situation in mind, balancing competing interests, recognizing what they have not said, and exercising discretion that honors both the letter and the spirit of the law.

The danger with AI, Lee argues, is that lawyers will be tempted to reduce their role to verifying machine output by emerging as syntactic workers in a semantic profession. Many jurisdictions legally distinguish between providing legal information, which almost anyone can do, and exercising independent legal judgment, which is the practice of law. This distinction is the profession's core claim to relevance; however, eroding it from within, by gradually reducing lawyers to editors of AI output, would be a costly self-inflicted wound.

Legal judgment is only half the answer

Even accepting Lee's framework, a long-standing blind spot remains. The legal profession uses one word, judgment, to describe two distinct cognitive capabilities, causing confusion around the use of the term.

Olga Mack, CEO of TermScout and a longtime builder of legal and decision systems, draws a sharp distinction between legal judgment and what she calls business judgment. Legal judgment produces lawyers who reason well about the law; while business judgment describes lawyers who can take that reasoning and translate it into something a business partner can understand and act upon. This can help the partner by prioritizing what matters, articulating trade-offs, mapping consequences, and handing the client something they can actually use. Without that translation layer, the legal work has limited effect.

While an AI tool can produce certain legal work faster, deciding which answers matter, understanding what a business needs to do with them, and taking responsibility for that advice is the human work that remains. Developing legal judgment while ignoring business judgment will produce lawyers who are only partially equipped for what comes next.

This is the first note of convergence. Lee, a former academic, and Mack, a legal executive, arriving independently at the same conclusion: The challenge is that the legal profession has not been clear enough about which human capabilities need to be preserved in the AI age, and why.

The professional judgment gap starts upstream

Both forms of judgment, legal and business, depend on something that is now under threat before lawyers ever reach a firm. Researchers studying AI's broader impact on professional formation have identified what they call a professional judgment gap in the hollowing out of the foundational experiences through which newer professionals develop critical thinking and decision-making skills.

The concern, as articulated by Dr. Bruna Damiana Heinsfeld of the University of Minnesota and Dr. Hidenori Tanaka, Group Leader of the Physics of AI Group at NTT Research, Inc. and the CBS-NTT Physics of Intelligence Program at Harvard University, is that academic institutions tend to adopt AI less out of pedagogical conviction than competitive anxiety. Being labeled AI ready has become a reputational imperative, regardless of whether the educational case for that designation has been made. The resulting consequence is a generation of graduates who may be able to work fluently with AI tools, but who have not developed the capacity to work independently of them. Like a calculator handed to a student before they understand arithmetic, AI introduced before foundational skills are established erodes the quality of judgment rather than enhancing it.

The timing compounds the problem. AI is being deployed most aggressively to automate the entry-level roles on which foundational professional skills have traditionally been built. Indeed, this is the work that teaches judgment through struggle, repetition, and the friction of making and fixing mistakes. Dr. Tanaka describes the likely result as a K-shaped cognitive economy in which experienced professionals tend to use AI to compound their existing advantages, while entry-level workers find their traditional pathways to expertise narrowing just as they come to rely on them. Left unaddressed, the gap between the two groups could widen over time.

That said, Dr. Tanaka frames the real divide as one of agency as much as seniority: senior professionals who stop staying hands-on and keeping up with the new tools can see their edge erode, while high-agency newcomers, less anchored by traditional assumptions and freer to invent new approaches, can leapfrog and even create entirely new kinds of roles and businesses.

Here the convergence deepens. Academic researchers with technology expertise, working from the literature on professional formation rather than from legal practice, have arrived at the same structural concern as Lee and Mack.

How some law firms are rethinking talent and leadership

Forward-thinking law firms are not waiting for the profession to resolve this at a systemic level. In fact, the choices they are making now reveal how the diagnosis is landing on the ground.

Norah Olson Bluvshtein, Chief Legal Operations Officer at Fredrikson & Byron, is asking how AI changes the way for which legal work gets paid and sees three possible futures for law firm business models: i) staying anchored to the billable hour while using AI to handle more volume; ii) shifting toward value-based pricing in which fixed fees align firm and client incentives more naturally; or  iii) a scenario in which frontier AI models become so capable that firms no longer need expensive legal-specific tools, fundamentally changing the cost structure under either pricing model

The honest answer, Olson Bluvshtein argues, is that the future probably holds some combination of all three. What is not optional, however, is treating technology strategy and people strategy as separate conversations, she explains, adding that the talent model has to be designed in tandem with the business model, even as uncertainty mounts about which model will prevail.

At Fredrikson, this has meant a three-part focus in associate development that works to:

• build AI fluency that goes beyond basic adoption;
• accelerates the development of legal judgment, including the ability to supervise and validate AI-produced work; and
• doubles down on the human skills that technology cannot replicate.

In recruiting, for example, the firm now prioritizes curiosity, demonstrated human capabilities, and openness over familiarity with specific tools.

Lorie Almon, Chair and Managing Partner of Seyfarth Shaw, pushes that argument further, reframing what leadership itself must become. Her starting point is an honest admission  about the apprenticeship model that has existed for decades but has never been examined closely. While the apprenticeship model produced capable lawyers for generations, the legal profession has long inferred causation from correlation. The assumption that lawyers learn judgment by muscling through  problems and accumulating instinct through repetition is  inferential — and there is no definitive proof about which inputs were essential and which were simply how things happened to be done.

This matters enormously right now, because many law firms are removing components of that developmental system before they fully understand which parts matter. The risk is that beautifully produced AI work product will create the illusion of competence while the underlying capacity for judgment quietly atrophies.

Almon's response is to reframe the role of firm leadership entirely. Leaders must become systems architects and design workflows that have AI supplementary components rather than replace effortful thinking. They also must build feedback loops that keep human mentorship at the center and create environments in which lawyers remain active participants in reasoning rather than passive editors of machine-generated output.

Deployed thoughtfully in this way, AI can provide feedback at a speed and frequency the traditional model never could, and early evidence of AI-enabled simulation tools has demonstrated this promise. The forward-thinking firms that will define the next era are already asking deeper questions of themselves, such as:

• What does it mean to be a great lawyer?
• Which human experiences are irreplaceable in building that greatness?
• How do we use AI as a catalyst for becoming even better at the craft?

The converging view around the need to break down the lawyer formation pipeline through deliberate redesign is well underscored by law firm executives working inside large law firms from entirely different vantage points, in addition to academic researchers with technology expertise and other legal experts. This converging view is a structural point the legal profession urgently needs to address.

Rebuilding the lawyer formation system

Individual firms adapting their traditional apprenticeship training programs is necessary but not sufficient. And that’s because the profession's deepest problem is that the apprenticeship model was never examined closely enough. Now that the model is being disrupted, the profession cannot simply defend it — mostly because it cannot fully describe it. Reform has to be partly a design project and partly a research project that can identify which experiences build judgment before deciding how to preserve or replace them.

Jordan Furlong, author of the this Substack, while arguing for a systemic overhaul of how lawyers’ skills are formed from law school through the early years of practice points out that the challenge is that law schools, legal employers, and state regulators each hold different pieces of the puzzle, and no single actor can solve it alone. Now, what has emerged from the research and conversations across this space is a framework of three interconnected pillars that can serve as an approach to modernization that is built on what is already understood.

The 3 pillars include:

1. Integrating work experience directly into legal education
2. Distilling legal judgment into teachable, defined micro-skills
3. Creating a phased, intentional approach to AI that treats it as a thinking partner throughout a lawyer's career

The first pillar is modeled on medical residency, meaning that the legal profession does not need to invent it from scratch. Several law schools have already pioneered versions of this approach. For example, Northeastern Law's Cooperative Legal Education Program guarantees every student nearly a full year of full-time work experience before graduation by alternating rigorous academic terms with full-time co-op placements across diverse legal environments. And UNH Franklin Pierce’s Daniel Webster Scholar Honors Program offers a competency-based bar exam alternative, developed in collaboration with the New Hampshire Supreme Court, that allows students to demonstrate readiness for practice through supervised experiential learning rather than a traditional two-day exam.

The second pillar revolves on the idea that the legal profession cannot teach what it cannot describe. Right now there is no agreed-upon definition of legal judgment, no shared standards for when and how AI should be used, and no structured curriculum for developing judgment deliberately. The components — which include pattern recognition, strategic calibration, reasoning through uncertainty, source evaluation, and ethical judgment under pressure — need to be named, defined, and taught rather than absorbed through osmosis.

In the third pillar, especially in the early stages, foundational legal reasoning should be developed without AI assistance to build analytical capability. Later, AI can enter as a research assistant with mandatory verification protocols. Advanced stages can involve immersive AI-integrated workflows used under proper supervision. The goal is to sequence the introduction of AI so that it accelerates development rather than substituting for it.

Where will AI take the legal profession?

What is most striking about this moment is not any single insight but the fact that so many legal experts across a multitude of disciplines are making the same point. Indeed, this convergence is itself an argument that suggests what is happening is not a niche concern of legal futurists but a structural shift that the entire profession needs to confront.

The five hours per week that AI is expected to return to lawyers represents the whole challenge. Those hours can be used to generate more syntactic volume, or they can be reinvested in the work that has given lawyers meaning for generations: Understanding clients, exercising judgment, fulfilling fiduciary duties, and maintaining the connection between law and human dignity.

The profession that stays anchored to the traditional ways of doing things and sleepwalks onto the first path will find itself automated into irrelevance from within. The one that chooses the second path will need to act with urgency by redesigning legal education, rebuilding the formation pipeline, aligning talent and technology strategies, and teaching judgment as a foundational competency that must be developed deliberately, early, and across their organizations and institutions.

What AI brings to the partnership is speed and breadth. What humans bring is judgment and the accountability that goes with it. This understanding is what will keep lawyers essential in the age of AI — but only if the profession chooses to preserve it.

You can find out more about the challenges AI poses to the legal profession here‍